记一次Flink通过Kafka写入MySQL的过程

梦痕长情

已于 2024-01-25 09:13:40 修改

阅读量1.4k

点赞数 21

文章标签： flink kafka mysql

于 2024-01-23 17:00:50 首次发布

本文链接：https://blog.csdn.net/m0_48830183/article/details/135631295

版权

一、前言
总体思路：source -->transform -->sink ,即从source获取相应的数据来源，然后进行数据转换，将数据从比较乱的格式，转换成我们需要的格式，转换处理后，然后进行sink功能，也就是将数据写入的相应的数据库DB中或者写入Hive的HDFS文件存储。
思路：
在这里插入图片描述

pom部分放到最后面。

二、方案及代码实现

2.1 Source部分
Source部分构建一个web对象用于保存数据等操作，代码如下：

package com.lzl.flink;

import java.util.Date;

/**
 * @author lzl
 * @create 2024-01-18 12:19
 * @name pojo
 */
public class Web {
    private String uuid;
    private String ip;
    private String area;
    private String web;
    private String operate;
    private Date createDate;

    public String getArea() {
        return area;
    }

    public String getIp() {
        return ip;
    }

    public String getOperate() {
        return operate;
    }

    public String getUuid() {
        return uuid;
    }

    public String getWeb() {
        return web;
    }

    public Date getCreateDate() {
        return createDate;
    }

    public void setArea(String area) {
        this.area = area;
    }

    public void setIp(String ip) {
        this.ip = ip;
    }

    public void setOperate(String operate) {
        this.operate = operate;
    }

    public void setUuid(String uuid) {
        this.uuid = uuid;
    }

    public void setWeb(String web) {
        this.web = web;
    }

    public void setCreateDate(Date createDate) {
        this.createDate = createDate;
    }
}

将生成的数据转化为JSON格式，测试如下：

public static void webDataProducer() throws Exception{
        //构建web对象，在ip为10.117后面加两个随机数
        int randomInt1 = RandomUtils.nextInt(1,255);
        int randomInt2 = RandomUtils.nextInt(1,999);
        int randomInt3 = RandomUtils.nextInt(1,99999);
        List<String> areas = Arrays.asList("深圳", "广州", "上海", "北京", "武汉", "合肥", "杭州", "南京");
        List<String> webs = Arrays.asList("www.taobao.com","www.baidu.com","www.jd.com","www.weibo.com","www.qq.com","www.weixin.com","www.360.com","www.lzl.com","www.xiaomi.com");
        List<String> operates = Arrays.asList("register","view","login","buy","click","comment","jump","care","collect");

        Web web = new Web();  //实例化一个web对象，并向对象中放入数据
        web.setUuid("uid_" + randomInt3);
        web.setIp("10.110." + randomInt1 +"." + randomInt2);
        web.setArea(getRandomElement(areas));
        web.setWeb(getRandomElement(webs));
        web.setOperate(getRandomElement(operates));
        web.setCreateDate(new Date());

        // 转换成JSON格式
        String webJson = JSON.toJSONString(web);
        System.out.println(webJson); //打印出来看看效果

    }
	//构建一个从列表里面任意筛选一个元素的函数方法
    public static <T> T getRandomElement(List<T> list) {
        Collections.shuffle(list);
        return list.get(0);
    }

    public static void main(String[] args) {
        while (true) {
            try {
                // 每三秒写一条数据
                TimeUnit.SECONDS.sleep(3);
                webDataProducer();
            } catch (Exception e) {
                e.printStackTrace();
            }
        }
    }

执行测试结果如下：
在这里插入图片描述
至此Source部分结束~~~~！！！！！！

2.2 Transform_1部分

2.2.1 写入kafka方法函数：

package com.lzl.flink;

import com.alibaba.fastjson.JSON;
import org.apache.commons.lang3.RandomUtils;
import org.apache.kafka.clients.producer.KafkaProducer;
import org.apache.kafka.clients.producer.ProducerRecord;

import java.util.*;
import java.util.concurrent.TimeUnit;

/**
 * @author lzl
 * @create 2024-01-18 12:18
 * @name KafkaWriter
 */
public class KafkaWriter {

    //kafka集群列表
    public static final String BROKER_LIST = "cdh39:9092,cdh40:9092,cdh41:9092";
    //kafka的topic
    public static final String TOPIC_WEB = "web";
    //kafka序列化的方式，采用字符串的形式
    public static final String KEY_SERIALIZER = "org.apache.kafka.common.serialization.StringSerializer";
    //value的序列化方式
    public static final String VALUE_SERIALIZER = "org.apache.kafka.common.serialization.StringSerializer";

    public static void writeToKafka() throws Exception {
        Properties props = new Properties(); //实例化一个Properties
        props.put("bootstrap.servers", BROKER_LIST);
        props.put("key.serializer", KEY_SERIALIZER);
        props.put("value.serializer", VALUE_SERIALIZER);

        // 构建Kafka生产者
        KafkaProducer<String, String> producer = new KafkaProducer<>(props);

        // 将web生成的数据发送给kafka的记录
        String webDataJson = webDataProducer();
        ProducerRecord<String,String> record = new ProducerRecord<String,String>(TOPIC_WEB,null,null,webDataJson);
        // 发送到缓存
        producer.send(record);
        System.out.println("向kafka发送数据：" + webDataJson);
        producer.flush();
    }
   public static void main(String[] args) {
        while (true) {
            try {
                // 每三秒写一条数据
                TimeUnit.SECONDS.sleep(3);
                writeToKafka();
            } catch (Exception e) {
                e.printStackTrace();
            }
        }
    }

2.2.2 建立 web的topic：
在这里插入图片描述
启动程序测试：

2.2.3 消费kafka看看是否有数据？

[root@cdh39 kafka]# bin/kafka-console-consumer.sh --bootstrap-server cdh39:9092 --from-beginning --topic web
{"area":"合肥","createDate":1705571020461,"ip":"10.110.104.676","operate":"comment","uuid":"uid_29661","web":"www.qq.com"}
{"area":"北京","createDate":1705571024048,"ip":"10.110.49.479","operate":"jump","uuid":"uid_77119","web":"www.weibo.com"}
{"area":"合肥","createDate":1705571027106,"ip":"10.110.232.960","operate":"click","uuid":"uid_99704","web":"www.taobao.com"}
{"area":"上海","createDate":1705571030140,"ip":"10.110.12.252","operate":"buy","uuid":"uid_99850","web":"www.jd.com"}
{"area":"合肥","createDate":1705571033228,"ip":"10.110.75.328","operate":"care","uuid":"uid_33135","web":"www.qq.com"}
{"area":"上海","createDate":1705571036267,"ip":"10.110.4.862","operate":"collect","uuid":"uid_37279","web":"www.taobao.com"}
{"area":"北京","createDate":1705571039361,"ip":"10.110.139.814","operate":"register","uuid":"uid_33016","web":"www.baidu.com"}
{"area":"武汉","createDate":1705571042422,"ip":"10.110.159.143","operate":"collect","uuid":"uid_26315","web":"www.lzl.com"}
{"area":"南京","createDate":1705571045495,"ip":"10.110.81.685","operate":"login","uuid":"uid_38712","web":"www.baidu.com"}
{"area":"南京","createDate":1705571048545,"ip":"10.110.228.267","operate":"comment","uuid":"uid_23297","web":"www.weibo.com"}
{"area":"武汉","createDate":1705571051623,"ip":"10.110.102.247","operate":"collect","uuid":"uid_77340","web":"www.lzl.com"}
{"area":"武汉","createDate":1705571054687,"ip":"10.110.184.832","operate":"comment","uuid":"uid_35230","web":"www.360.com"}
{"area":"武汉","createDate":1705571057760,"ip":"10.110.90.361","operate":"buy","uuid":"uid_52082","web":"www.lzl.com"}
{"area":"北京","createDate":1705571060825,"ip":"10.110.37.707","operate":"buy","uuid":"uid_45343","web":"www.weixin.com"}
{"area":"上海","createDate":1705571063909,"ip":"10.110.178.901","operate":"care","uuid":"uid_51015","web":"www.baidu.com"}
{"area":"杭州","createDate":1705571066945,"ip":"10.110.153.758","operate":"collect","uuid":"uid_46772","web":"www.xiaomi.com"}
{"area":"合肥","createDate":1705571069980,"ip":"10.110.177.755","operate":"comment","uuid":"uid_78442","web":"www.taobao.com"}
{"area":"广州","createDate":1705571073020,"ip":"10.110.151.427","operate":"register","uuid":"uid_92174","web":"www.weixin.com"}
{"area":"上海","createDate":1705571076072,"ip":"10.110.217.622","operate":"jump","uuid":"uid_86059","web":"www.xiaomi.com"}

至此，Transform_1部分结束~~~！！！！

2.3 Sink部分
创建一个MySQLSink，继承RichSinkFunction类。重载里边的open、invoke 、close方法，在执行数据sink之前先执行open方法，然后开始调用invoke方法，调用完之后最后执行close方法关闭资源。即在open里面创建数据库连接，然后调用invoke执行具体的数据库写入程序，完毕之后调用close关闭和释放资源。这里要继承flink的RichSinkFunction接口。代码如下：

package com.lzl.flink;

import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.functions.sink.RichSinkFunction;

import java.sql.Connection;
import java.sql.PreparedStatement;
import java.sql.SQLException;
import java.sql.Timestamp;
import java.util.logging.Logger;

/**
 * @author lzl
 * @create 2024-01-22 15:30
 * @name MySqlToPojoSink
 */
public class MySqlToPojoSink extends RichSinkFunction<Web> {

    private static final Logger log = Logger.getLogger(MySqlToPojoSink.class.getName());
    private static final long serialVersionUID = 1L;
    private Connection connection = null;
    private PreparedStatement ps = null;
    private String tableName = "web";
    
 	@Override
    public void open(Configuration parameters) throws Exception {
        super.open(parameters);
        log.info("获取数据库连接");

        // 通过Druid获取数据库连接，准备写入数据库
        connection = DbUtils.getConnection();

        // 插入数据库的语句   因为我们封装的pojo的类型为PojoType<com.lzl.flink.Web, fields = [area: String, createDate: Date, ip: String, operate: String, uuid: String, web: String]>
        String insertQuery = "INSERT INTO " +   tableName + "(time,ip,uid,area,web,operate) VALUES (?,?,?,?,?,?)" ;

        // 执行插入语句
        ps = connection.prepareStatement(insertQuery);
    }

	 // 重新关闭方法。   关闭并释放资源
    @Override
    public void close() throws Exception {
    super.close();
        if(connection != null) {
            connection.close();
        }
        if (ps != null ) {
            ps.close();
        }
    }

 	// 重写invoke方法
    @Override
    public void invoke(Web value,Context context) throws Exception {
        //组装数据，执行插入操作
        ps.setTimestamp(1, new Timestamp(value.getCreateDate().getTime()));
        ps.setString(2,value.getIp());
        ps.setString(3, value.getUuid());
        ps.setString(4, value.getArea());
        ps.setString(5, value.getWeb());
        ps.setString(6, value.getOperate());
        ps.addBatch();

        // 一次性写入
        int[] count = ps.executeBatch();
        System.out.println("成功写入MySQL数量：" + count.length);
    }
}

特别说明：从kafka读取到的内容是String，里面包含JSON格式。本文是先将它封装成Pojo对象，然后在Sink这里解析它的Value。（开始是尝试通过apply算子将它转换为List，但是失败了（时间有限，后续再弄），最后是通过map算子）

至此，Sink部分结束~！

2.4 Transform_2部分。消费kafka 数据，添加Sink。

package com.lzl.flink;

import com.alibaba.fastjson.JSON;
import org.apache.flink.api.common.serialization.SimpleStringSchema;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.connectors.kafka.FlinkKafkaConsumer;
import java.util.Properties;
import java.util.concurrent.TimeUnit;
/**
 * @author lzl
 * @create 2024-01-19 8:49
 * @name DataSourceFromKafka
 */
public class DataSourceFromKafka {
    public static void transformFromKafka() throws Exception {
        // 构建流执行环境
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setParallelism(1);
		//kafka 配置
        Properties prop = new Properties();
        prop.put("bootstrap.servers", KafkaWriter.BROKER_LIST);
        prop.put("zookeeper.connect", "cdh39:2181");
        prop.put("group.id", KafkaWriter.TOPIC_WEB);
        prop.put("key.serializer", KafkaWriter.KEY_SERIALIZER);
        prop.put("value.serializer", KafkaWriter.VALUE_SERIALIZER);
        prop.put("auto.offset.reset", "earliest");

        // 建立流数据源
        DataStreamSource<String> dataStreamSource = env.addSource(new FlinkKafkaConsumer<String>(
                KafkaWriter.TOPIC_WEB,
                new SimpleStringSchema(),
                prop
        )).setParallelism(1); // 单线程打印，控制台不乱序，不影响结果

        SingleOutputStreamOperator<Web> webStream = env.addSource(new FlinkKafkaConsumer<>(
                "web",
                new SimpleStringSchema(),
                prop
        )).setParallelism(1)
                .map(string-> JSON.parseObject(string,Web.class));

        webStream.addSink(new MySqlToPojoSink());
        env.execute();
    }
    
    public static void main(String[] args) throws Exception {
        while (true) {
            try {
                // 每1毫秒写一条数据
                TimeUnit.MILLISECONDS.sleep(1);
                transformFromKafka();
            } catch (Exception e) {
                e.printStackTrace();
            }
        }
    }
}

如果要设置空值报错异常，或者排除空值可以：

package com.lzl.flink;

import com.alibaba.fastjson.JSON;
import org.apache.flink.api.common.serialization.SimpleStringSchema;
import org.apache.flink.streaming.api.CheckpointingMode;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.environment.CheckpointConfig;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.functions.ProcessFunction;
import org.apache.flink.streaming.connectors.kafka.FlinkKafkaConsumer;
import org.apache.flink.util.Collector;
import java.util.Properties;
import java.util.concurrent.TimeUnit;

public class DataSourceFromKafka {


    public static void transformFromKafka() throws Exception {
        // 构建流执行环境
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setParallelism(1);

        //checkpoint设置
        //每隔10s进行启动一个检查点【设置checkpoint的周期】
        env.enableCheckpointing(10000);
        //设置模式为：exactly_one，仅一次语义
        env.getCheckpointConfig().setCheckpointingMode(CheckpointingMode.EXACTLY_ONCE);
        //确保检查点之间有1s的时间间隔【checkpoint最小间隔】
        env.getCheckpointConfig().setMinPauseBetweenCheckpoints(1000);
        //检查点必须在10s之内完成，或者被丢弃【checkpoint超时时间】
        env.getCheckpointConfig().setCheckpointTimeout(10000);
        //同一时间只允许进行一次检查点
        env.getCheckpointConfig().setMaxConcurrentCheckpoints(1);
        //表示一旦Flink程序被cancel后，会保留checkpoint数据，以便根据实际需要恢复到指定的checkpoint
        env.getCheckpointConfig().enableExternalizedCheckpoints(CheckpointConfig.ExternalizedCheckpointCleanup.RETAIN_ON_CANCELLATION);
        //设置statebackend,将检查点保存在hdfs上面，默认保存在内存中。先保存到resources目录下
        env.setStateBackend(new FsStateBackend("D:java//Flink1.17//src//main//resources"));


//         kafka 配置
        Properties prop = new Properties();
        prop.put("bootstrap.servers", KafkaWriter.BROKER_LIST);
        prop.put("zookeeper.connect", "cdh39:2181");
        prop.put("group.id", KafkaWriter.TOPIC_WEB);
        prop.put("key.serializer", KafkaWriter.KEY_SERIALIZER);
        prop.put("value.serializer", KafkaWriter.VALUE_SERIALIZER);
        prop.put("auto.offset.reset", "earliest")

        DataStreamSource<String> webStream = env.addSource(new FlinkKafkaConsumer<>(
                "web",
                new SimpleStringSchema(),
                prop
        )).setParallelism(1);

        //使用process算子 排除空值
        DataStream<Web> processData = webStream.process(new ProcessFunction<String, Web>() {
            @Override
            public void processElement(String s, Context context, Collector<Web> collector) throws Exception {
                try {
                    Web webs = JSON.parseObject(s, Web.class);
                    if (webs != null) {
                        collector.collect(webs);
                    }
                } catch (Exception e) {
                    System.out.println("有空值数据");
                }
            }
        });

        processData.addSink(new MySqlToPojoSink());
        env.execute();
    }

    public static void main(String[] args) throws Exception {
        while (true) {
            try {
                // 每1毫秒写一条数据
                TimeUnit.MILLISECONDS.sleep(1);
                transformFromKafka();
            } catch (Exception e) {
                e.printStackTrace();
            }
        }
    }
}

至此，Transfrom结束~！

2.5 DB部分（这部分可以先做，或者放到前面，因为需要测试）
本次的DB演示采用常规的MySQL数据库。采用Druid工具连接。
思路：创建一个数据库连接的工具，用于连接数据库。使用Druid工具，然后放入具体的Driver，Url，数据库用户名和密码，初始化连接数，最大活动连接数，最小空闲连接数也就是数据库连接池，创建好之后返回需要的连接。

package com.lzl.flink;

import com.alibaba.druid.pool.DruidDataSource;
import java.sql.Connection;
/**
 * @author lzl
 * @create 2024-01-18 17:58
 * @name DbUtils
 */
public class DbUtils {
    private static DruidDataSource dataSource;

    public static Connection getConnection() throws Exception {
        dataSource = new DruidDataSource();
        dataSource.setDriverClassName("com.mysql.cj.jdbc.Driver");
        dataSource.setUrl("jdbc:mysql://cdh129:3306/flink?useSSL=true");
        dataSource.setUsername("root");
        dataSource.setPassword("xxb@5196");
        // 设置初始化连接数，最大连接数，最小闲置数
        dataSource.setInitialSize(10);
        dataSource.setMaxActive(50);
        dataSource.setMinIdle(5);
        // 返回连接
        return dataSource.getConnection();
    }
}

数据库建表语句：

CREATE TABLE `web_traffic_analysis` (
  `time` varchar(64) DEFAULT NULL COMMENT '时间',
  `ip` varchar(32) DEFAULT NULL COMMENT 'ip地址',
  `uid` varchar(32) DEFAULT NULL COMMENT 'uuid',
  `area` varchar(32) DEFAULT NULL COMMENT '地区',
  `web` varchar(64) DEFAULT NULL COMMENT '网址',
  `operate` varchar(32) DEFAULT NULL COMMENT '操作'
) ENGINE=InnoDB DEFAULT CHARSET=utf8 ROW_FORMAT=DYNAMIC COMMENT='网页流量分析表'

三、启动程序
开始本来是想将上面所有的功能都写成函数方法，然后单独开一个Main（）主函数的入口，然后在主函数下面调用那些方法（生产数据、消费数据方法）。思路是借鉴python的：if name == ‘main’: 下调用很多的方法。但实际执行过程，是先生成数据，然后将数据写入kafka，然后再消费数据，过程执行非常慢，这个方案被pass了。后来又想到多线程方案，一个线程跑生产数据和写入数据，一个线程跑消费数据和写入下游数据库。这个方法是测试成功了，但是跑了一会儿就出现数据的积压和内存oom了，因为我设定的是1毫秒生产一条数据，写入kafka也需要一定的时间，加上电脑内存不足，有点卡，这个方案也被pass了。最后的方案是将生产数据打包放到集群去跑，本地电脑开启消费kafka数据写入MySQL数据库。结果如下：
生产数据：
在这里插入图片描述
消费和写入数据库数据：

数据库数据：

至此结束，后面有其他想法再补充~！

多线程部分代码：

package com.example.study;

import com.lzl.flink.DataSourceFromKafka;
import com.lzl.flink.KafkaWriter;

public class WebApplication {
    public static void main(String[] args) throws Exception {
    // 创建线程1
     Thread threadOne = new Thread(new Runnable() {
        @Override
        public void run() {
            while (true){
                try{
                    KafkaWriter kafkaWriter = new KafkaWriter();
                    kafkaWriter.webDataProducer();
                    kafkaWriter.writeToKafka();
                    System.out.println("线程一在跑~！");
                } catch (Exception e) {
                    e.printStackTrace();
                }
            }
        }
    });

  // 创建线程2
 Thread threadTwo = new Thread(new Runnable() {
        @Override
        public void run() {
            while (true) {
                DataSourceFromKafka dataSourceFromKafka = new DataSourceFromKafka();
                try {
                    dataSourceFromKafka.transformFromKafka();
                    System.out.println("线程二在跑~！");
                } catch (Exception e) {
                    e.printStackTrace();
                }
            }
        }
    });

    //启动线程
    threadOne.start();
    threadTwo.start();
    Thread.sleep(5);

    }
}

结果：
在这里插入图片描述

梦痕长情

关注

21
点赞
踩
29

收藏

觉得还不错? 一键收藏
0
评论
记一次Flink通过Kafka写入MySQL的过程

这个方法是测试成功了，但是跑了一会儿就出现数据的积压和内存oom了，因为我设定的是1毫秒生产一条数据，写入kafka也需要一定的时间，加上电脑内存不足，有点卡，这个方案也被pass了。总体思路：source -->transform -->sink ,即从source获取相应的数据来源，然后进行数据转换，将数据从比较乱的格式，转换成我们需要的格式，转换处理后，然后进行sink功能，也就是将数据写入的相应的数据库DB中或者写入Hive的HDFS文件存储。至此，Transform_1部分结束~~~！
复制链接

扫一扫